A0516
Title: Predicting the popularity of tweets using internal and external knowledge: An empirical Bayes approach
Authors: Feng Chen - UNSW Syd (Australia) [presenting]
Wai Hong Tan - UNSW Sydney (Australia)
Abstract: The problem of tweet popularity prediction is considered. We model the retweet time sequence using an inhomogeneous Poisson process with the intensity function depending on the age of the original tweet and the calendar time, through a parametric and a nonparametric function respectively. The functional parameter is estimated nonparametrically using the training data, and the finite-dimensional parameter using both internal knowledge on the times of historical retweets up to the censoring time, and external knowledge on complete retweet sequences in the training data set. The internal and external knowledge are then combined using a novel empirical Bayes type approach, where the prior distribution for the model parameter is constructed based on the external knowledge, and the likelihood is calculated based on the internal knowledge. The mode of the posterior distribution is then used as the estimator of the finite-dimensional parameter. Suitable functionals of the predictive distribution for the number of retweets implied by the estimated model are then used to predict the tweet popularity. The proposed methodology was applied to a large Twitter data set, and its performance is found to be superior to those of the competing approaches in the literature.
